The short video platform, which has over 600 million active users, announced the new tool on June 6. It’s called Kling. Like OpenAI’s Sora model, Kling is able to create videos “up to two minutes at 30 fps frame rate and up to 1080p video resolution.” the company says on its website.
But unlike Sora, which still remains unavailable to the public four months after testing OpenAI, Kling soon began letting people test the model themselves.
I was one of them. I accessed it after downloading Kuaishou’s video editor, registering with a Chinese number, joining a waiting list, and filling out an additional form through Kuaishou’s user feedback groups. The model cannot process prompts written entirely in English, but you can overcome this by either translating the phrase you want to use into Chinese or including a Chinese word or two.
Well, first things first. Here are some results I created with Kling to show you what it’s like. Remember Sora’s impressive demo video with scenes on the streets of Tokyo or the cat runs through a garden? Here are Kling’s downloads:
ZEYI YANG/MIT TECHNOLOGY OVERVIEW | KLING
ZEYI YANG/MIT TECHNOLOGY OVERVIEW | KLING
ZEYI YANG/MIT TECHNOLOGY OVERVIEW | KLING
Remember the image of Dall-E’s equestrian astronaut? I asked Kling to create a video version as well.
ZEYI YANG/MIT TECHNOLOGY OVERVIEW | KLING
There are a few things worth applauding here. None of these videos stray too far from the prompt, and the physics look right—the camera panning, the ruffled leaves, and the way the horse and astronaut turn, showing Earth behind them. The production process took about three minutes for each of them. Not the fastest, but perfectly acceptable.
But there are also obvious shortcomings. The videos, while 720p in format, look blurry and grainy. Sometimes Kling ignores an important request in the prompt. And most importantly, all videos created are now limited to five seconds, which makes them much less dynamic or complex.
However, it’s not really fair to compare these results to things like Sora’s demos, which have been selected by OpenAI to be released to the public and probably represent better than average results. These Kling videos are some of the first attempts I made with each prompt and I rarely included direct engineering keywords like “8k, photorealism” to refine the results.